Proper node order for grain, NR, sharpening and 709 LUT

I've been trying to find a good tutorial on this but have struggled. There is plenty on ordering nodes for basic adjustments things like primaries, contrast, correcting vs grading etc. But I am interested in understanding the proper placement of the following:
Noise Reduction
Sharpening
Rec709 LUT
Grain
And not only where they would go individually in the node tree, but also in relation to each other.

Currently I put a mild sharpening node at the end of the tree in Group-Post-Clip (because I'm shooting BRAW, so they all need it), and individually sharpen clips if needed.
Noise reduction at the beginning of the tree on each clip individually.
Rec709 LUT, I used to apply to the clip directly in the media pool, but found bad clipping issues until I started applying it to the beginning of the node tree on the clip instead (i'd be interested in an explanation of why). Does applying to the clip in the media pool apply the LUT before or after the colour tab's node trees?
Grain, I figure goes at the very end, after sharpening.

My general structure

Log to DWG CST > Noise Reduction > WB > Exposure > Color > DWG to 709 CST > Film emulation

- NR goes near the end for me, so i can use as little of it as posiable, just won't know till the grade is near complete how little you can get away with, so at the end for me, and as little as liveable
- rarely / never touch sharpening, if we are really in trouble it gets sent to VFX, there is sharpening options in debayer settings for Venice /Alexa /Red, and 98% of what i see is shot one of those camera's, i can tweak in raw settings if needed, i default to 0 on all camera's as they are all a bit too sharp with good glass
- i work in ACES or RCM2, so no use / need for / want of a display LUT in my workflow, in both RCM and ACES tone mapping to a display is last in the chain

- NR goes near the end for me, so i can use as little of it as posiable, just won't know till the grade is near complete how little you can get away with, so at the end for me, and as little as liveable

100% agree, don't turn on NR until you see what if any noise you have, exposure set.

This doesn't mean the node has to be at the end, you can put it early and just leave it off. Certainly better for performance to leave it off while you do other adjustments.

That said, i have seen recommendations both ways, put NR node early, or late. I would be interested to see if anyone could generate A/B test to prove one or the other is 'better'.

Thanks everyone.
My question was less about the order in which to make adjustments, and more about the order in which to place the nodes, i know this matters with some and less with others, depending on the adjustment in question. I do generally put the NR close to the beginning of the node tree, but it sits as an empty node until everything else is done. I only actually do NR near the end.
My workflow is usually: LUT - Exposure - WB - Local adjustments - look - sharpening - NR - Grain.
But my node tree is: NR - Exposure - WB - Locals - sharp - LUT - grain

I need fairly quick turnarounds, so I use The One LUT for the BMPCC4k. I am not familiar with this log to DWG workflow. Sounds like it's worth some experimentation, perhaps. I've experimented with CST to RCM2 and tone mapping, but I'm not skilled enough to get good results. The One LUT is beautiful on the BM4k.

Sharpening is necessary. I use a speed booster XL and often shoot 120p, so it's not the sharpest or cleanest image. Lots of NR at times, because of the low light, often in forrest, and 120p. Hence more need to sharpen

That said, i have seen recommendations both ways, put NR node early, or late. I would be interested to see if anyone could generate A/B test to prove one or the other is 'better'.

I don't think in terms of pure NR only there can be a lot of difference but workflow wise it does matter. People tend to recommend doing it as the very first step before any other operation to give the grading pipe a much cleaner signal. This matters most if you like to use qualifiers a lot. You should of course judge the noise by viewing the image all the way through your grade or at least normalized to display to judge the amount needed. But you need to be aware of what part of the noise might be due to heavy grading and what actually comes from the source. NR is easily overdone if you're not considerate.

I do generally put the NR close to the beginning of the node tree, but it sits as an empty node until everything else is done. I only actually do NR near the end.
My workflow is usually: LUT - Exposure - WB - Local adjustments - look - sharpening - NR - Grain.
But my node tree is: NR - Exposure - WB - Locals - sharp - LUT - grain

So yea this approach is totally sensible as long as you turn of your grain before dialing in NR.

1. Noise Reduction
2. Sharpening
3. Rec709 LUT
4. Grain

I would get rid of 2 and 4, replace 3 with RCM and put 1 at the end.

1. Noise Reduction
2. Sharpening
3. Rec709 LUT
4. Grain

I would get rid of 2 and 4, replace 3 with RCM and put 1 at the end.

Can you explain your rational for NR at end, even after grain?
And can you explain your rational for RCM?
I assume you're talking about a CST to RCM, correct?

I was actually advocating not using grain or sharpening.

For RCM I was actually advocating using Color Management, so not a CST.

First I would suggest using color management... properly.

What I do generally for node order (proper for me anyways):
NR - after the noise and before any corrections, especially those which could be impacted by noise. So if to reduce camera noise, it will be at beginning of node tree. If to reduce noise of an operation then immediately after that node (although I would likely avoid any operation causing noise or artifacts in the first place.) Question to keep in mind: why grade the noise rather than suppressed noise? Some grading will give different results of the noise if before, after or parallel.

Sharp and grain - after other corrections because these need to be uniform over the image or will really show as their effects can be neutralized by other grading. Grain would be after sharp. Again depends on what grading might matter.

All the above could be off during the grade so as not to bog down the system.

any LUT - last after grading, but I would suggest if grain used, then grain would be last if the LUT messes with the grain. CAUTION: whatever the LUT, it must be created to be used with the correct working color space. (input and output of the LUT must both be compatible with the space it is within.) Best place for a LUT is in the timeline portion of the node tree... last unless grain or unless if node based color management transform (which is then usually last.) Note that a LUT should be the second thing turned on after color management so that any grading is done "under" the LUT. Reason for all this is that a LUT will remap according to a matrix of values to force a certain look according to certain assumed in and out color space. Working after the LUT fights against it. And working in a different color space from what the LUT was designed for just dosn't work.

Also, (and as I think others may have mentioned) if only using a LUT to transform to rec.709 it is far better to just do that with the color management. Difference of math vs guessing points of a matrix. That said, LUTs can be a nice shortcut to get a certain created look.

If you are shooting at higher resolution and your hardware is up to it, I’d even suggest doing NR at source resolution. Most of our sources are compressed these days and scaling may ‘clump’ pixel groups in an ugly way, hard to treat by NR.

Regarding grain there's no technically correct way to apply it because it is a creative choice in post anyway. But there can be two ideas to follow.

1. Grain early in the tree after denoise and management to simulate film stock grain. When grading film you also grade the grain along with it. The downside is this becomes tricky when you like to use qualifiers during grading because you added noise. Same reason why NR should go first which is why I find Jim's suggestion to do it at the end very impractical.

2. Grain post grade but pre display conversion. To simulate film print grain which would be applied after the grade. I say pre display conversion because in my opinion no operator should be post transform because it makes those decisions display dependent/display-referred. You'd need to re-evaluate them should you go to a different output intent.